You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For sake of proper API design and to simplify unit testing a cluster node should be able to be started and stopped over and over again.
Currently, a node's API is asymmetrical in that resources are "started" in constructor and may be stopped explicitly using method stop(). However, shell exposes method start() for starting current node, too. Due to the lack of support for starting a node separately, this is working once, only.
The text was updated successfully, but these errors were encountered:
For sake of proper API design and to simplify unit testing a cluster node should be able to be started and stopped over and over again.
Currently, a node's API is asymmetrical in that resources are "started" in constructor and may be stopped explicitly using method
stop()
. However, shell exposes methodstart()
for starting current node, too. Due to the lack of support for starting a node separately, this is working once, only.The text was updated successfully, but these errors were encountered: